Trust, explained

The Trust Layer for AI Agents

Agents do not need “more tools.” They need tools they can trust. FunctionFly turns functions into verified, signed, auditable building blocks with execution-backed trust scores.

Verification levels Trust API

Verification levels

Each level adds deeper assurance before a tool can be considered “trusted.”

L1
Format checks
Validate manifest structure and I/O schema.
L2
Security scans
Scan for risky behaviors and validate capability constraints.
L3
Code review
Manual and automated review of safety-relevant aspects.
L4
Platform verified
Signed and reviewed as official/recommended tooling.

Signing, attestations, revocation

Trust is portable because verification becomes an immutable record.

  • Signing: Function artifacts include platform signatures.
  • Attestations: Verifiable records of what was checked and when.
  • Revocation: If a tool is flagged, it can be downgraded or removed from trusted pools.
What agents see
A trust score and policy-relevant metadata (capabilities, constraints, and verification level).

Execution-backed trust

Trust scores compound from real execution history so “it ran” and “it ran safely” are both represented.

Zero-knowledge vault

Secrets are encrypted client-side. FunctionFly stores ciphertext only, enabling agent tools without exposing plaintext secrets.

Trust API (usage-based)

Query attestations, trust scores, and revocation state. Bill on usage so agent workloads can adapt trust dynamically.

Trust score lifecycle

Trust is not a static label. It starts at publish-time, evolves with execution history, and can be updated through verification, reports, and revocations.

1
Publish
A function is published with a manifest and a declared I/O contract.
2
Verify
Verification stages validate structure, security, and safety-relevant behavior.
3
Sign & attest
Attestations become immutable records that agents can reason about.
4
Execute
Real execution updates trust score components over time.
5
Evolve
History, reports, and revocation keep trust current for agents and clients.

Trust API for agents

Use a scoped API key to read trust scores and trigger verification workflows from your agent runtime.

  • Get trust score: GET /v1/trust/score/{function_id}
  • Batch scores: POST /v1/trust/batch
  • Get history: GET /v1/trust/history/{function_id}
  • Submit verification: POST /v1/trust/verify
Scopes matter
Trust read endpoints require an API key; verification and reporting endpoints require the corresponding scopes (e.g. verification:request and reports:submit).

Example: submit verification

Trigger a verification run for a function version, then query status until the verification completes. 

POST /v1/trust/verify
Authorization: Bearer API_KEY
Content-Type: application/json

{
  "function_id": "550e8400-e29b-41d4-a716-446655440002",
  "function_version": "1.2.0",
  "verification_level": "standard",
  "metadata": {
    "use_case": "data_processing",
    "expected_traffic": "high"
  }
}
{
  "id": "550e8400-e29b-41d4-a716-446655440005",
  "verification_id": "vfy_abc123def456",
  "function_id": "550e8400-e29b-41d4-a716-446655440002",
  "function_author": "alice",
  "function_name": "data-processor",
  "function_version": "1.2.0",
  "verification_level": "standard",
  "status": "pending",
  "created_at": "2026-03-21T10:00:00Z"
}

Why FunctionFly wins on trust

No other platform combines verification, trust scoring, and zero-knowledge security.

Feature
FunctionFly
RapidAPI
Toolhouse
Theagora
Trust scoring
✓
—
—
Partial
Multi-level verification
✓
—
—
✓
Zero-knowledge vault
✓
—
—
—
Function execution
✓
—
—
✓
Agent-native tool discovery
✓
—
✓
✓

Ready to build with trusted tools?

Browse trusted functions in the marketplace, or run your own agent trust policy to select tools safely.

Get Started Free See trust-aligned pricing